Mosquitoes or Snakes — which scares you more?

Countries with Risk of Dengue

(Based on 2019 data)

Why are countries near the equator at higher risk of dengue?”

Image Source: https://www.gptx.org/Departments/Public-Health-and-Environmental-Quality/Public-Health/Mosquito-Control-Fight-the-Bite/Mosquito-Facts

Favorable climate that speed up the mosquito life cycle

  1. Warm temperatures year-round

  2. High humidity and frequent rainfall

  3. No harsh winters

  4. Many equatorial countries have rapidly growing urban areas with poor waste and water management.

  5. People often live and work in open environments with limited protection, increasing exposure.

Countries with Risk of Dengue

Sri Lanka

Time Series Plot of Weekly Dengue Cases in Sri Lanka (National Level)

Time Series Plot of Weekly Dengue Cases in Sri Lanka (National Level)

Red Segment: Interrupted Period Due to the COVID-19 Pandemic

Data

Weekly Dengue Cases Corresponds to 25 Districts in Sri Lanka

Data Source

denguedatahub R package

On CRAN

install.packages("denguedatahub")

You could install the development version from Github using

install.packages("devtools")
devtools::install_github("thiyangt/denguedatahub")

Web scrapping using the denguedatahub R package

More about denguedatahub

link: https://denguedatahub.netlify.app/

Methodology: Methods of Forecasting

Methods of Addressing Interrupted Period

Approach 1

Excluding the interrupted period from model training

Approach 2

Set interrupted period to missing

Approach 3

Forecasting the interrupted period first and then using it for modeling

Approach 4

Down-weighting observations

Approach 4

Set interrupted period to missing

Comparison

  • Forecasts are compared without adjusting for the interruption period

Excluding the interrupted period from model training

Data

training period:

test period:

Method

  • ARIMA models through auto.arima: without applying log transformation on cases

  • ARIMA models through auto.arima: with applying log transformation on cases (bias-corrected back transformation for the mean prediction)